Pregnancy outcomes before and after institution of a specialised twins clinic: a retrospective cohort study

Our study demonstrates that multidisciplinary care in a specialised twins clinic was associated with significant differences in maternal and neonatal outcomes compared to conventional models of care. Maternal outcome differences included reduction in Caesarean section rate and reduced total maternal inpatient stay. Regarding fetal outcomes, reduction in late prematurity was significant. These differences were achieved without any apparent increase in maternal or fetal complications.

Although Caesarean rates were higher in both ANC and private vs. TC, the reasons appeared to differ. Given similar rates of labour and induction/augmentation in ANC vs. TC, increased ANC intrapartum Caesarean rates likely accounted for the difference. Regarding TC and private women, more TC women underwent labour, with the difference between these cohorts probably due to increased private elective Caesarean rates. We cannot exclude factors such as a higher rate of non-vertex first twins in private women (with consequent Caesarean section), however it is unlikely this would account for the entire observed 19 % difference. Importantly, many of these Private elective Caesareans were occurring late preterm (34 + 0–36 + 6 weeks), with the late preterm and term Caesarean rates being 82 and 63 % respectively for Private women (Vs. 42 and 54 % respectively in TC women). The increased late preterm elective (predominantly Caesarean) birth in turn partially accounts for increased neonatal nursery admission rates in the Private cohort. However, 28 + 0–33 + 6 week preterm birth was also more common in the Private cohort (and at that gestation Caesarean rate was approximately 80 % in all groups). Although it is not possible from this study to obtain reasons for the high rate of late preterm elective Caesareans in the Private cohort, it may relate to lower tolerance of risk, particularly as regards MCDA twins.

This study was undertaken prior to publication of the Twin Birth trial [13], and TC clinicians discussed with women the available epidemiological evidence about perinatal outcomes and mode of birth [14, 15]. These standardised discussions and clear documentation of delivery plan may have increased confidence of TC women and staff regarding intrapartum care, resulting in decreased intrapartum Caesarean rate compared to ANC. Certainly substantially more vaginal deliveries for TC women involved a vaginal breech birth (11 % vs. 1 % ANC, p < 0.001), possibly reflecting greater acceptance of second twin vaginal breech birth in this group.

Although modest on an individual basis, the one day reduction of total pregnancy maternal inpatient stay in TC compared to ANC has considerable health system economic implications. The average Australian cost of a pregnancy/childbirth admission in 2011 (length of stay 2.5 days) was AUD 4815 (GBP 3100.86, USD 4978.00, EUR 3572.99 on average 2011 interbank rate) [16], so a consistently shorter stay represents substantial savings. Clearly it is not ascertainable from the study databases the reason for the length of stay reduction, although it was found both slightly shorter antepartum admissions and slightly shorter postpartum stay contributed. We speculate that the specialised TC care led to greater confidence amongst women and clinicians regarding appropriate earlier discharge for both antenatal admissions and the birth admission. This is supported by the smaller interquartile range for length of stay in TC compared to ANC.

We found a significant reduction in late preterm birth in TC. As high risk twin pregnancies were triaged to MFM care, and there were no differences in antenatal complications, the reason for this difference in prematurity is unclear. It may be due to greater confidence in methods of fetal surveillance while awaiting term delivery in the TC cohort. Given emerging concerns regarding consequences of late prematurity, we speculate that TC care may confer important long term benefits [17, 18].

Study strengths include the reasonably large cohort, known chorionicity derived in high proportion from ultrasound reports and placental pathology, and use of the private patient groups to help overcome any historical cohort effect.

The limitations of this study include its retrospective and non-randomised nature, as differences in outcome might be secondary to differences in patient characteristics, and/or confounders such as general improvement in outcomes over time. Regarding underlying characteristics, there were no significant TC and ANC demographic differences, and although private women were slightly older, maternal complications such as hypertension did not differ between groups. Additionally, chorionicity did not differ between TC and private, making it less likely that the outcome differences noted can be explained by different risk profiles between groups.

Regarding time-related versus care-related findings, analysis was performed both of private controls being seen in the TC era, and of private women being seen January 2007 to March 2009 (pre-TC) versus April 2009 to November 2011. Demographic characteristics of pre-TC and TC private women were similar apart from the later cohort being older. No substantive differences in private women’s maternal or fetal outcomes were noted over time. This makes it less likely that the change in outcomes that occurred after TC introduction reflects only a general time-related improvement in outcomes.

As model of care was not randomised, selection bias is also a potential limitation. Although all women booked with ANC were offered transfer to TC when it commenced, and TC booking was the default model of care for public bookings after March 2009, 11 of 84 ANC women had their babies after TC established (of whom 7 were already beyond 20 weeks’ gestation when TC commenced). As retrospective note review does not reveal reasons for women staying with ANC, selection bias is possible: however, demographic factors and chorionicity did not differ between these women and the total ANC cohort, and the initial 2009 TC women (n = 15) were if anything of higher risk profile than subsequent TC women (53 % Vs. 27 % MCDA, p = 0.07), so unlikely to bias findings in favour of TC. The decision was therefore made to include all ANC and TC women in the outcome analysis, without excluding the crossover time period.

The choice of which models of care to include could also be questioned. This was based on an attempt to compare populations of broadly equivalent risk i.e. ANC, TC and Private. As women triaged to MFM <20 weeks would usually be because of a likely fetal complication, and women transferred in from another hospital would usually be secondary to complications such as premature labour, these populations were considered inherently high-risk and not an appropriate comparison. Conversely, a small proportion of low-risk women were looked after in midwifery-led care (permitted for DCDA twins if the woman had booked into this model of care, wished to remain in this model of care, and had no major maternal conditions or initial known obstetric complications), so were considered of lower baseline risk and not appropriate for direct comparison. As there were only 18 such women (6 pre-TC, 12 in TC era) it is unlikely their exclusion has substantially altered the overall study results: however, is a further reason why a randomised trial would be the preferred method of demonstrating utility of a multiple pregnancy clinic Vs. standard care.

As for all retrospective studies, both accuracy and scope of database-derived data may be a limitation. As fields such as chorionicity, model of care, perinatal mortality, and maternal inpatient stay were checked against multiple sources, and fields relating to labour and birth have been shown to be accurate in prior local research [19], we are confident of the data pertaining to our primary and major secondary outcomes. However, some data appear contradictory: in particular, the rate of threatened premature labour (TPL) was lower in Private patients than TC or ANC while other markers of threatened or actual preterm birth (e.g. steroid administration, gestation at birth) are higher. As the “threatened premature labour” field may encompass both admitted and non-admitted cases, has yes/no/unsure input rather than more detailed information, and might be more specifically asked about/known for public rather than private patients by the midwives filling in the database, we believe this result likely reflects the limitations of the database rather than a true difference in TPL rates between groups.

Further limitations include the lack of formal patient satisfaction information (although TC women did report favourably on the convenience of having ultrasound and visit scheduled at the same time), or formal economic analysis. Both of these would ideally be addressed using a prospective, randomised design in a centre where equipoise regarding TC care still exists.

Prior comparable studies are summarized in Table 4 [6‐8, 20]. In the one published RCT (162 women) of specialised antenatal care for multiple pregnancy, the additional midwifery care and information trialled did not improve maternal or fetal morbidities [20]. However, this trial was underpowered for such outcomes and was not a trial of specialised obstetric antenatal care. Other studies regarding the value of specialist multiple pregnancy care are, like this study, retrospective in nature, and all performed in the USA [21]. Outcomes vary, but all are broadly supportive of improvement in fetal and/or maternal outcomes, although none found decreased Caesarean rate. However, all also either included only DCDA twins, or had no chorionicity information available. Given the importance of chorionicity in determining twin pregnancy outcome [9], lack of this information makes interpretation of these studies especially problematic.

Although the value of decreased length of stay, decreased rates of prematurity, and decreased nursery admission rates found in this study are widely accepted, whether a decrease in Caesarean rates for twin pregnancy is desirable has been contested, given epidemiological data regarding increased perinatal morbidity and mortality for the second-born twin [22, 23]. However, Barrett et al’s recent large RCT found no significant difference in fetal, neonatal or maternal outcomes between elective Caesarean and planned vaginal delivery for twins ≥32 weeks where Twin 1 was vertex presentation [13]. Caesarean section has recognized implications for subsequent pregnancies and is generally more costly to the healthcare system than vaginal birth [24]. Therefore our observed reduction in Caesarean deliveries, without an increase in adverse events, should indeed be viewed as a positive outcome.

Although our study supports the establishment of specialised Twins Clinics, further research is required. An economic analysis of the establishment of a Twins Clinic, examining whether reduced inpatient stay, prematurity, and Caesarean rates offset clinic implementation costs (including staff education and ultrasound service provision), is essential for long-term viability of this model of care. While one prior study found a substantial decrease in immediate neonatal costs after specialised antenatal care for twin pregnancy was introduced, neither maternal cost of care nor implementation costs were assessed [7], and we are not aware of any prior studies where a full economic evaluation of a specialised multiple pregnancy clinic has been performed. Ultimately, a sufficiently powered, randomised controlled trial of Twins Clinic versus other models of care will be needed to confirm the benefits identified in studies such as ours.